Sequential recruitment of accumbal dopamine and acetylcholine during learning

Abstract

Associative learning depends on neuromodulator release in the nucleus accumbens, including both dopamine and acetylcholine. Dopamine signals convey prediction errors that drive learning and dynamically evolve as predictions are acquired, but the role of acetylcholine in learning is far less clear. We hypothesized that acetylcholine signals change over the course of learning, complementing dopaminergic encoding of prediction errors. To test this hypothesis, we examined how acetylcholine and dopamine dynamics co-evolve during learning by recording these two signals as rats learned three cue-guided operant tasks. We discovered that neuromodulatory responses to task cues evolve according to a specific sequential structure and correlate with different features of associative learning. When rats learn a cue-outcome association, there is first the appearance of indiscriminate cholinergic dips, then dopamine responses differentiate based on predicted value, followed by cholinergic dips also segregating according to value. Stable performance required this full pattern, and acetylcholine dips could be decoupled from value by changing reward contingencies. Moreover, optogenetic manipulation of accumbal cholinergic interneurons during cue presentation bidirectionally altered learning rate, with activation slowing and inhibition facilitating associative learning. Using computational modeling, we demonstrate that the observed acetylcholine dynamics best fit an associative salience term predicted by hybrid attentional associative learning models, suggesting that dopamine and acetylcholine act complementarily to determine the content and rate of learning.